tf_upgrade_v2 on resnet and utils folders. #6154

goldiegadde · 2019-02-04T21:57:30Z

No description provided.

* Add resnet56 short tests. - created base benchmark module - renamed accuracy test class to contain the word Accuracy which will result in a need to update all the jobs and a loss of history but is worth it. - short tests are mostly copied from shining with oss refactor * Address feedback. * Move flag_methods to init - Address setting default flags repeatedly. * Rename accuracy tests. * Lint errors resolved. * fix model_dir set to flags.data_dir. * fixed not fulling pulling out flag_methods.

* Add short imagenet tests (taken from seemuch) - also rename to match go forward naming * fix method name * Update doc strings. * Fixe gpu number.

Failed test is python2 and was a kokoro failure

* Add short imagenet tests (taken from seemuch) - also rename to match go forward naming * fix method name * Update doc strings. * Fixe gpu number. * Add fill_objects * fixed calling wrong class in super. * fix lint issue.

* Fix the turn_off_ds flag problem * add param names to all args

) * Export benchmark stats using tf.test.Benchmark.report_benchmark() * Fix python style using pyformat

tfboyd · 2019-02-05T03:29:45Z

@guptapriya This is LGTM to me and I think we power through the lint errors separately as the majority are over 80 charcters per line issues die to compat.v1. This also bit rots really fast. What do you think? I ran some tests and TF 2.0 worked. If the TF 1.0 CIFAR-10 test that @goldiegadde did was 92.6 to 93% then I say we go with it.

… gadde-v2-convert

goldiegadde · 2019-02-05T06:29:29Z

@guptapriya This is LGTM to me and I think we power through the lint errors separately as the majority are over 80 charcters per line issues die to compat.v1. This also bit rots really fast. What do you think? I ran some tests and TF 2.0 worked. If the TF 1.0 CIFAR-10 test that @goldiegadde did was 92.6 to 93% then I say we go with it.

@tfboyd @guptapriya
Updated the PR to fix all the lint errors. I did run the cifar benchmark tests, will update the result tomorrow morning.

guptapriya

some comments, but most likely we will want to address these in subsequent PRs.

guptapriya · 2019-02-05T06:19:18Z

official/resnet/cifar10_download_and_extract.py

@@ -60,4 +60,4 @@ def _progress(count, block_size, total_size):

 if __name__ == '__main__':
 FLAGS, unparsed = parser.parse_known_args()
- tf.app.run(argv=[sys.argv[0]] + unparsed)
+ tf.compat.v1.app.run(argv=[sys.argv[0]] + unparsed)


can't we use absl`s app.run here? (simiar to cifar10_main.py)?

We can ask and then see if they can fix it. We can also manually change it.

guptapriya · 2019-02-05T06:37:36Z

official/resnet/cifar10_main.py

@@ -273,6 +274,6 @@ def main(_):


 if __name__ == '__main__':
- tf.logging.set_verbosity(tf.logging.INFO)
+ tf.compat.v1.logging.set_verbosity(tf.compat.v1.logging.INFO)


I was checking the docs and I think we are supposed to use absl for logging in 2.0?
https://github.com/tensorflow/docs/blob/master/site/en/r2/guide/effective_tf2.md#api-cleanup

Looks like the upgrade script doesn't do this automatically. Can we do this in a subsequent PR perhaps?

from absl import logging logging.info(...) logging.error(...)

https://abseil.io/docs/python/guides/logging

Yes, this is on our list. I was shocked that no logging moves forward in TF 2.0 without a rewrite.

guptapriya · 2019-02-05T06:43:05Z

official/resnet/keras/keras_cifar_main.py

@@ -81,7 +81,7 @@ def parse_record_keras(raw_record, is_training, dtype):
 Tuple with processed image tensor and one-hot-encoded label tensor.
 """
 image, label = cifar_main.parse_record(raw_record, is_training, dtype)
- label = tf.sparse_to_dense(label, (cifar_main.NUM_CLASSES,), 1)
+ label = tf.compat.v1.sparse_to_dense(label, (cifar_main.NUM_CLASSES,), 1)


can you file a ticket to fix this? We should not need to do this anyway and it might be better to just figure out the right way to do this.

We are going to go back and file tickets on anything compat.v1 in the keras code path.

guptapriya · 2019-02-05T06:44:25Z

official/resnet/keras/keras_cifar_main.py

@@ -98,7 +98,7 @@ def run(flags_obj):
 Dictionary of training and eval stats.
 """
 if flags_obj.enable_eager:
- tf.enable_eager_execution()
+ tf.compat.v1.enable_eager_execution()


i think maybe we should also add an else clause here for the opposite case so it can work in 2.0 too?

else: tf.compat.v1.disable_eager_execution()

wdyt?

We will make a 2.0 branch as some things are in 2.0 that cannot be done in 1.0 like saving a model in 2.0 format, the call just does not exist.

guptapriya · 2019-02-05T06:44:50Z

official/resnet/keras/keras_imagenet_main.py

@@ -88,7 +88,7 @@ def run(flags_obj):
 ValueError: If fp16 is passed as it is not currently supported.
 """
 if flags_obj.enable_eager:
- tf.enable_eager_execution()
+ tf.compat.v1.enable_eager_execution()


same here re: disable eager in else case

guptapriya · 2019-02-05T06:46:27Z

official/resnet/layer_test.py

@@ -72,10 +72,10 @@ def _batch_norm_ops(self, test=False):

 g = tf.Graph()
 with g.as_default():
- tf.set_random_seed(self.name_to_seed(name))
- input_tensor = tf.get_variable(
+ tf.compat.v1.set_random_seed(self.name_to_seed(name))


i think in 2.0 this is tf.random.set_seed
https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/framework/random_seed.py#L187

guptapriya · 2019-02-05T06:52:07Z

official/resnet/layer_test.py

- tf.set_random_seed(self.name_to_seed(name))
- input_tensor = tf.get_variable(
+ tf.compat.v1.set_random_seed(self.name_to_seed(name))
+ input_tensor = tf.compat.v1.get_variable(


is get_variable supposed to be just tf.Variable?

guptapriya · 2019-02-05T06:56:01Z

official/resnet/resnet_run_loop.py

@@ -88,7 +88,7 @@ def process_record_dataset(dataset,

 # Parses the raw records into images and labels.
 dataset = dataset.apply(
- tf.contrib.data.map_and_batch(
+ tf.data.experimental.map_and_batch(


Sigh. yeah this file will need a lot of work to make it 2.0 friendly

guptapriya · 2019-02-05T06:56:54Z

official/resnet/resnet_run_loop.py

- targets=labels,
- k=5,
- name='top_5_op'))
+ accuracy = tf.compat.v1.metrics.accuracy(labels, predictions['classes'])


i believe most of these metrics and summaries can use their newer versions, we just need to do it carefully and make sure everything works fine.

tfboyd · 2019-02-05T15:53:14Z

I suspect we may (maybe maybe not) need some copybara changes but we can deal with that when we try to import to Google3 post merge. Maybe we get lucky and all it works out.

tfboyd · 2019-02-05T15:53:48Z

Merging now to avoid future conflicts and we have all agreed to address the comments with bugs as needed over the next 48 hours.

* Add resnet56 short tests. (tensorflow#6101) * Add resnet56 short tests. - created base benchmark module - renamed accuracy test class to contain the word Accuracy which will result in a need to update all the jobs and a loss of history but is worth it. - short tests are mostly copied from shining with oss refactor * Address feedback. * Move flag_methods to init - Address setting default flags repeatedly. * Rename accuracy tests. * Lint errors resolved. * fix model_dir set to flags.data_dir. * fixed not fulling pulling out flag_methods. * Use core mirrored strategy in official models (tensorflow#6126) * Imagenet short tests (tensorflow#6132) * Add short imagenet tests (taken from seemuch) - also rename to match go forward naming * fix method name * Update doc strings. * Fixe gpu number. * points default data_dir to child folder. (tensorflow#6131) Failed test is python2 and was a kokoro failure * Imagenet short tests (tensorflow#6136) * Add short imagenet tests (taken from seemuch) - also rename to match go forward naming * fix method name * Update doc strings. * Fixe gpu number. * Add fill_objects * fixed calling wrong class in super. * fix lint issue. * Flag (tensorflow#6121) * Fix the turn_off_ds flag problem * add param names to all args * Export benchmark stats using tf.test.Benchmark.report_benchmark() (tensorflow#6103) * Export benchmark stats using tf.test.Benchmark.report_benchmark() * Fix python style using pyformat * Typos. (tensorflow#6120) * log verbosity=2 logs every epoch no progress bars (tensorflow#6142) * tf_upgrade_v2 on resnet and utils folder. * tf_upgrade_v2 on resnet and utils folder.

This reverts commit d6b2b83.

…6162)" (#6167) This reverts commit 57e0752.

tfboyd and others added 10 commits February 4, 2019 12:15

Use core mirrored strategy in official models (#6126)

ce73f81

Imagenet short tests (#6132)

39b1273

* Add short imagenet tests (taken from seemuch) - also rename to match go forward naming * fix method name * Update doc strings. * Fixe gpu number.

points default data_dir to child folder. (#6131)

3ab2869

Failed test is python2 and was a kokoro failure

Imagenet short tests (#6136)

c4ece8a

* Add short imagenet tests (taken from seemuch) - also rename to match go forward naming * fix method name * Update doc strings. * Fixe gpu number. * Add fill_objects * fixed calling wrong class in super. * fix lint issue.

Flag (#6121)

17a777d

* Fix the turn_off_ds flag problem * add param names to all args

Export benchmark stats using tf.test.Benchmark.report_benchmark() (#6103

e57a082

) * Export benchmark stats using tf.test.Benchmark.report_benchmark() * Fix python style using pyformat

Typos. (#6120)

9a4f074

log verbosity=2 logs every epoch no progress bars (#6142)

d1f5d1a

tf_upgrade_v2 on resnet and utils folder.

8b18b5b

goldiegadde requested review from karmel and a team as code owners February 4, 2019 21:57

googlebot added the cla: yes label Feb 4, 2019

Merge branch 'master' into gadde-v2-convert

69ce0a2

tfboyd requested review from tfboyd and removed request for a team and karmel February 4, 2019 22:12

tfboyd approved these changes Feb 5, 2019

View reviewed changes

goldiegadde added 4 commits February 4, 2019 21:35

tf_upgrade_v2 on resnet and utils folder.

682736c

Merge branch 'gadde-v2-convert' of github.com:goldiegadde/models into…

020e374

… gadde-v2-convert

Merge branch 'gadde-v2-convert' of github.com:goldiegadde/models into…

bab39b4

… gadde-v2-convert

Merge branch 'gadde-v2-convert' of github.com:goldiegadde/models into…

9ca3ce7

… gadde-v2-convert

guptapriya approved these changes Feb 5, 2019

View reviewed changes

tfboyd merged commit d6b2b83 into tensorflow:master Feb 5, 2019

goldiegadde deleted the gadde-v2-convert branch February 6, 2019 19:18

tfboyd pushed a commit that referenced this pull request Feb 6, 2019

Revert "tf_upgrade_v2 on resnet and utils folders. (#6154)" (#6162)

57e0752

This reverts commit d6b2b83.

tfboyd pushed a commit that referenced this pull request Feb 8, 2019

Revert "Revert "tf_upgrade_v2 on resnet and utils folders. (#6154)" (#…

b2c9e3f

…6162)" (#6167) This reverts commit 57e0752.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

tf_upgrade_v2 on resnet and utils folders. #6154

tf_upgrade_v2 on resnet and utils folders. #6154

goldiegadde commented Feb 4, 2019

tfboyd commented Feb 5, 2019

goldiegadde commented Feb 5, 2019

guptapriya left a comment

guptapriya Feb 5, 2019

tfboyd Feb 5, 2019

guptapriya Feb 5, 2019

tfboyd Feb 5, 2019

guptapriya Feb 5, 2019

tfboyd Feb 5, 2019

guptapriya Feb 5, 2019

tfboyd Feb 5, 2019

guptapriya Feb 5, 2019

guptapriya Feb 5, 2019

guptapriya Feb 5, 2019

guptapriya Feb 5, 2019

guptapriya Feb 5, 2019

tfboyd commented Feb 5, 2019

tfboyd commented Feb 5, 2019

tf_upgrade_v2 on resnet and utils folders. #6154

tf_upgrade_v2 on resnet and utils folders. #6154

Conversation

goldiegadde commented Feb 4, 2019

tfboyd commented Feb 5, 2019

goldiegadde commented Feb 5, 2019

guptapriya left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

tfboyd commented Feb 5, 2019

tfboyd commented Feb 5, 2019